Duffee (2005) shows that the amount of consumption risk (i.e., the conditional covariance between market returns and consumption growth) is procyclical. In light of this “Duffee Puzzle,” I empirically demonstrate that the conditional covariance between dividend growth (i.e., the immediate cash flow part of market returns) and consumption growth is (1) procyclical and (2) a consistent source of procyclicality in the puzzle. Moreover, I solve an external habit formation model that incorporates realistic joint dynamics of dividend growth and consumption growth. The procyclical dividend-consumption comovement entails two new procyclical terms in the amount of consumption risk via cash flow and valuation channels, respectively. These two procyclical terms play an important role in generating a realistic magnitude of consumption risk. In contrast to extant habit formation models, the conditional equity premium no longer increases monotonically when a negative consumption shock arrives because it might lower the amount of risk while increasing the price of risk.

We propose a fiscal policy expectations mechanism. When bad macro news arrives (in our study, when initial jobless claims (IJC) are higher than expected), investors may expect more generous government spending and drive up aggregate stock prices through the expected cash flow channel. Using a time-series sample from January 2013 to March 2021, we find that this phenomenon emerges when newspapers mention fiscal policy more. In the cross section, firms expected to receive more government spending – through stimulus supports during COVID-19 or procurement contracts before 2020 – exhibit higher individual stock returns when bad IJC shocks arrive.